Open AccessSelection of Influential Genetic Markers Among a Large Number of Candidates Based on Effect Estimation Rather than Hypothesis Testing
Author(s) -
Ulf Strömberg,
Jonas Björk,
Karin Broberg,
Fredrik Mertens,
Paolo Víneis
Publication year - 2008
Publication title -
epidemiology
Language(s) - English
Resource type - Journals
eISSN - 1531-5487
pISSN - 1044-3983
DOI - 10.1097/ede.0b013e3181632c3d
Subject(s) - estimation , selection (genetic algorithm) , multiple comparisons problem , statistical hypothesis testing , genetic association , statistics , biology , single nucleotide polymorphism , computer science , genetics , econometrics , machine learning , mathematics , genotype , gene , management , economics
In epidemiologic studies on direct genetic associations, hypothesis testing is primarily considered for evaluating the effects of each candidate genetic marker, eg, single nucleotide polymorphisms. To help investigators protect themselves from over-interpreting statistically significant findings that are not likely to signify a true effect-a problem connected to multiple comparisons-consideration of the false-positive report probability has been proposed. There have also been arguments advocating estimation of effect size rather than hypothesis testing (P value). Here, we propose an estimation-based approach that offers an attractive alternative to the test-based false-positive report probability, when the task is to select promising genetic markers for further analyses. We discuss the potential of this estimation-based approach for genome-wide association studies.